Any discussion of the prior art throughout the specification should in no way be considered as an admission that such prior art is widely known or forms part of the common general knowledge in the field.
Optical coherence tomography (OCT) is a widely used interferometric technique for studying biological samples including in vivo tissue such as the human eye, with lateral and depth resolution, using information contained within the amplitude and phase of reflected or scattered light. OCT systems generally utilise a Michelson interferometer configuration, with two main approaches being employed: time domain OCT and spectral domain OCT.
In time domain OCT coherence properties of a partially coherent source such as a superluminescent light emitting diode (SLED) with a coherence length of several microns are utilised by interfering light reflected from a sample with a reference beam provided by the same source, but with a time-varying path length. At a specific depth in the sample corresponding to the path length delay in the reference arm, an interference envelope of fringes will be detected in the combined back-reflected signal, allowing the reflection profile in the depth dimension to be reconstructed. Commonly this is done for only a single sample point at a time, and the corresponding scan of depth is known as an ‘A-scan’.
Instead of scanning a delay line, spectral domain OCT techniques analyse the reflected light by interfering it with a reference beam, either as a time-varying function of wavelength (swept source OCT) or by dispersing the different wavelengths with a grating or other spectral demultiplexer and detecting them simultaneously along a detector array. The spectral domain information is the Fourier transform of the spatial (depth) reflection profile, so the spatial profile can be recovered by a Fast Fourier Transform (FFT). Generally speaking, spectral domain OCT systems are preferred over time domain OCT systems because they have a ˜20 to 30 dB sensitivity advantage.
OCT techniques can be adapted to provide a laterally resolved ‘B-scan’ by scanning the sample beam relative to the sample in one axis, or a ‘C-scan’ by scanning in two axes. Faster acquisition is generally desirable irrespective of the type of scan, especially for reducing motion-induced artefacts with in vivo samples, and has been greatly improved over the previous 20 to 25 years by advances in several fields including faster swept source scanning rates and photodetector array readout speeds. However a fundamental limitation with scanning spot schemes, especially for in vivo applications, is presented by laser safety regulations: reducing dwell time to increase scanning speed without being able to increase the applied power will inevitably degrade the signal to noise ratio.
Consequently there has also been research into ‘parallelised’ OCT systems in which an extended sample area is probed with lateral resolution, or an array of sample spots probed simultaneously. It is relatively straightforward to parallelise time domain OCT, e.g. by utilising a CCD camera and imaging optics as described in U.S. Pat. No. 5,465,147 entitled ‘Method and apparatus for acquiring images using a CCD detector array and no transverse scanner’. This provides a two dimensional (2-D) en face image, with depth resolution provided by scanning the reference mirror as usual in time domain OCT.
Swept source spectral domain OCT can be parallelised in similar fashion, as described in Bonin et al ‘In vivo Fourier-domain full-field OCT of the human retina with 1.5 million A-lines/s’, Optics Letters 35(20), 3432-3434 (2010). However because each frame corresponds to a single wavelength, the acquisition time for each A scan is equal to the frame period times the number of k-points (wavelength samples) acquired. Even for very high speed cameras with frame rates of 100s of kHz, this can lead to A scan acquisition times of many ms which can lead to motion artefacts especially with in vivo samples. PCT patent application No PCT/AU2015/050788, entitled ‘Multichannel optical receivers’, discloses an alternative parallelised swept source OCT scheme that enables faster acquisition. In one particular implementation a plurality of spots on a sample are illuminated simultaneously and the reflected or scattered signal light mixed with a reference beam to form a plurality of interferograms with unique carrier frequencies.
Parallelised spectrometer-based spectral domain OCT enables single shot B-scan acquisition, although existing schemes are limited by the fact that one axis of a 2-D photodetector array is occupied by the wavelength dispersion. In a configuration described in published US patent application No 2014/0028974 A1 entitled ‘Line-field holoscopy’, cylindrical lenses are used to produce a line illumination on a sample and on a reference mirror. As shown schematically in FIG. 1, the combined return sample and reference beams from a line illumination 2 are dispersed with a dispersive element such as a grating 4 and detected with a 2-D sensor array 6. A Fourier transform along the spectral axis 8 provides an A-scan for each position 9 along the illuminated line 2. For full three-dimensional (3-D) imaging the illuminated line is mechanically scanned in the orthogonal direction and the 2-D sensor array read out repeatedly.
Even if a linear B-scan of a sample is sufficient, i.e. 3-D imaging isn't required, a scan in the orthogonal direction may still be necessary, e.g. for digital wavefront correction to correct for lens aberrations and the like, or to provide increased depth of field. Furthermore for these purposes the repeated linear scans have to be phase coherent, which is generally difficult.
It is generally preferred for spectral domain OCT apparatus to be configured to sample the unambiguous complex field of the interference signal, rather than just the detected real-valued interference signal, to distinguish positive and negative path length delays and therefore enable imaging over the full depth of field range. A variety of approaches for capturing the complex field have been described. For example Jungwirth et al ‘Extended in vivo anterior eye-segment imaging with full-range complex spectral domain optical coherence tomography’, Journal of Biomedical Optics 14(5), 050501 (2009) describes, for a scanning spot scheme, a solution in which the sample phase is dithered as the sample is scanned. A key drawback of this approach is that sample movement can cause loss of phase coherence during scanning. Line field systems, which have improved phase stability, have been described which do not require dithering of the sample phase. In US 2014/0028974 A1 for example the complex field is obtained by sampling the signal in the far field of a linear illumination, whilst in Huang et al ‘Full-range parallel Fourier-domain optical coherence tomography using a spatial carrier frequency’, Applied Optics 52(5), 958-965 (2013), the line field is captured in the image plane, with an off-axis reference providing access to the complex field.
The transverse resolution of an OCT apparatus is determined, for a given wavelength, by the numerical aperture of the objective lens. However increasing the numerical aperture of the objective invariably reduces the depth of field, resulting in a trade-off between transverse resolution and depth of field. A variety of software-based or digital focusing techniques have been proposed to overcome this trade-off to increase the depth of field. These approaches generally assume that the phase coherence between scattering points is maintained during scanning and sample collection, and the field may be captured in the image plane or the Fourier plane.
In one example, synthetic aperture techniques are discussed in Mo et al ‘Depth-encoded synthetic aperture optical coherence tomography of biological tissues with extended focal depth’, Optics Express 23(4), 4935-4945 (2015). In another example, the forward model (FM) approach of Kumar et al ‘Numerical focusing methods for full field OCT: a comparison based on a common signal model’, Optics Express 22(13), 16061-16078 (2014), involves sampling the 3-D captured interferometric signal I(x, y, k) in the image plane using a full field swept source OCT apparatus with a 2-D CMOS camera. An unambiguous phase is obtained by requiring the sample to be on one side only of the zero delay, and the defocus correction is achieved by applying a numerical phase correction based on a Fresnel wavefront propagation model. This numerical phase correction is achieved by first performing a 1-D FFT of the real valued signal along the spectral axis to give the complex field, I(x, y, k)→E(x, y, Δz). This is followed by a 2-D FFT of the lateral coordinates for all positive delays, E(x, y, Δz)→E(kx, ky, Δz). The Fresnel correction for defocus correction is then applied: E(kx, ky, Δz)→E(kx, ky, Δz)γ,
where
  γ  =            exp      (              i        ⁢                                  ⁢                                            λ              0                        ⁢            Δ            ⁢                                                  ⁢                          zM              2                                            4            ⁢            π            ⁢                                                  ⁢            n                          ⁢                  (                                    k              x              2                        +                          k              y              2                                )                    )        .  Here, the wavelength is replaced by the centre wavelength λ0, n is the refractive index of the sample and M is the magnification of the OCT apparatus. A 2-D inverse FFT (IFFT) with respect to the spatial frequencies of the phase-corrected field gives an image focused over the full volume.
Digital focusing with a full-range line-field OCT system has been demonstrated in Fechtig et al ‘Full range line-field parallel swept source imaging utilizing digital refocusing’, Journal of Modern Optics (2014), DOI: 10.1080/09500340.2014.990938. In this case the sample field is measured in the image plane and full range measurements are achieved by using an off-axis configuration of the reference arm. This off-axis configuration introduces a lateral carrier frequency which shifts the interference term in frequency space enabling the positive and negative frequency components to be separated, thereby enabling measurement of the complex signal. Phase noise in the scanning direction restricts the digital focusing to one dimension, which is applied to each successive B scan. The complex signal is obtained by first taking a 1-D FFT along the spatial axis corresponding to the off-axis reference, after which a filter can be applied to select the positive frequency signal component from its complex conjugate artefact and the non-interferometric background. A 1-D IFFT then gives a signal measurement with unambiguous phase. Digital focusing is achieved by performing a 1-D FFT along the spectral axis followed by a 1-D FFT of the lateral coordinates to give E(kx, Δz), where Δz now extends over the full range. Multiplication by the 1-D phase correction factor followed by a 1-D IFFT gives the focused B-scan over the full range.
A full-field swept source OCT system with sampling in the far field is described in Hillmann et al ‘Holoscopy—holographic optical coherence tomography’, Optics Letters 36(13), 2390-2392 (2011). In this system, 2-D interferograms for each wavelength are propagated to a specific delay Δz. A 1-D FFT along the spectral axis is then used to reconstruct the focused object for this depth Δz. This process is repeated for a range of delays and the refocused regions are then stitched together. Full range imaging with sampling in the Fourier plane has been demonstrated using an off-axis reference beam to obtain an unambiguous phase, as described in Hillmann et al ‘Efficient holoscopy image reconstruction’, Optics Express 20(19), 21247-21263 (2012). This numerical post-processing approach, in which the 3-D signal is interpolated onto a non-equally spaced grid, provides a volume image with a resolution equivalent to the focal plane resolution throughout an extended portion of the volume. A final 3-D FFT then gives the focused volume image. Similar methods are used in inverse synthetic aperture microscopy (ISAM), described for example in Ralston et al ‘Interferometric synthetic aperture microscopy’, Nature Physics 3(2), 129-134 (2007).
We note that the approaches described above assume a simple model for depth-dependent defocus. An alternative approach that compensates for unknown optical aberrations using sub-aperture correlations is described in Kumar et al ‘Subaperture correlation based digital adaptive optics for full field optical coherence tomography’, Optics Express 21(9), 10850-10866 (2013).
An important limitation of full-field OCT systems, compared to point-scanning systems, is that that they are susceptible to crosstalk from multi-path scattering and hence have reduced sensitivity. In addition, the lack of confocal filtering increases the susceptibility to spurious reflections from outside the coherence length of the system. The line field approach of US 2014/0028974 A1 partially alleviates these limitations compared to that of a full field system by confocal gating in one axis. An alternative approach to mitigating crosstalk is to use a spatially incoherent source.
Unless the context clearly requires otherwise, throughout the description and the claims the words ‘comprising’, ‘comprises’ and the like are to be construed in an inclusive sense as opposed to an exclusive or exhaustive sense. That is, they are to be construed in the sense of ‘including, but not limited to’.